Indexer: Break down the values length and tag count metrics by metric type

From discussing with a few folks, it seems that the values length (in partiuclar) and tags count metrics graphs are not clear because they are not broken down by metric type. I think we should surface this as another tag to make it immediately clear whether the oncall is experiencing a situation of "tons of counters" vs. "some very very large distributions" when diagnosing an increase in bytes/message size.